Load Balancing ESXi 6.0 to Procurve

[zillion08]

Hi Everyone,

I'm trying to do some load balancing on my 5406zl switch for my ESXi 6.0 server.

Ports that connect to ESXI are D1-D8. D2-D8 are the non-management ports, and the ports I want to trunk.

Currently, where all servers on the ESXi are able to talk, the configuration looks something like this.:

---

Vlan 1

tagged D2-D8

IP address 10.1.0.1 255.255.0.0

Vlan 2

untagged D2-D8

IP address 10.2.0.1 255.255.0.0

---

I've tried doing:

Trunk D2-D8 trk16 trunk

and adding trk16 tagged to Vlan 1 and untagged to Vlan 2.

No dice.

---

Now I did notice that the server didn't save the NIC teaming on ESXi to be Route Based on IP hash, so I will try this tomorrow morning. But I'm not sure if that would cause all servers to not be able to communciate, when moving from non-trunk to trunk.

Any help would be greatly appreciated!

[Vince-Whirlwind · Accepted solution]

1/ You should not pass any untagged VLAN across inter-switch links. (If your hardware doesn't support not having an untagged VLAN, just use a null VLAN.)

2/ ESX used to always default to a hash method that didn't work - you used to have to change it to something different. I would have thought they would have fixed that by now, but maybe not.

3/ Otherwise, just as you have done: create the trunk, then assign the trunk as tagged to whichever VLANs need to cross it. Then mirror that config on the ESX side of the link. 
Personally, I would leave the working ports exactly as is, configure two new ports into a trunk, then re-patch the ESX server and configure it appropriately for the new ports. Makes your rollback slightly easier. Also helps you start your new trunks with a clean config using ports that don't have any existing config on them. Once you have a trunk working on those ports, add in the additional ports.

[zillion08]

@Vince-Whirlwind: The tagging and untagging makes since and I will attempt to do this tomorrow morning. I won't be able to move ports as the switch is full everywhere else. But why/how are my servers able to communicate now?

And, does this look correct? 

trunk D2-D8 trk16 trunk

vlan 1
name "DEFAULT_VLAN"
no untagged A1-A2,A15-A18,Trk49
untagged A21-A24,B1-B11,D10-D24,Trk1-Trk6,Trk9-Trk11,Trk13-Trk15
tagged A3-A14,D1, trk16
ip address 10.1.0.1 255.255.0.0
ip helper-address 10.2.1.21
exit
vlan 2
name "Building Data"
untagged A2-A18,D1,Trk49
tagged Trk2-Trk6, Trk16
ip address 10.2.0.1 255.255.0.0
ip helper-address 10.2.1.21
ip igmp
exit
vlan 200
name "Phones"
tagged trk16
ip address 10.200.0.1 255.255.0.0
ip helper-address 10.2.1.21
ip igmp
exit

Thanks again for your help!

[zillion08]

I beleive I have figured out why the servers are able to talk.

If you look here it says that VLAN ID 0, is ESXi, disables vlan tagging. Most of my servers are, in ESXi, are part of VLAN 0 instead of VLAN 1, 2, etc. (In my defense, I went off of what my predecessor did).

Because the server ports on the router (D2-D8) are untagged, everything can talk.

Now time to go and add proper vlans on ESXi.

I will update tomorrow morning if everything is working.

[zillion08]

It worked!

I made sure to tag traffic on ESXi, which I think why it wasn't working to begin with.

All vlans are tagged with the trunk.

When I made the change, on ESXi, to load balance based on hash, I had to restart all the VMs.

Thank you!!!